In this paper I apply an old problem of Quine's (the inscrutability of reference in translation) to a new style of theory about mental content (causal/nomological/informational accounts of meaning) and conclude that no “naturalization” of content of the sort currently popular can solve Quine's “gavagai” enigma. I show how failure to solve the problem leads to absurd conclusions not about one's own mental life, but about the nonmental world. I discuss various ways of attempting to remedy the accounts so as to avoid the problem and explain why each attempt at solving the problem would take the information theorists further from their self-assigned task of “naturalizing” semantics.